Skip to Main content Skip to Navigation

Estimation algorithms for thermoacoustic instabilities with distributed and nonlinear dynamics

Abstract : Unwanted thermoacoustic instabilities are harmful to combustion systems that suffer from them such as gas turbine combustors operating under lean premixed conditions. Advanced monitoring systems are needed to estimate and forecast the phenomenon to assist in decision making and automatic stabilization. In this Thesis we propose using a distributed description of acoustics interfaced to heat release models, with nonlinearities whenever possible, to describe the instabilities. State and parameter estimation algorithms taking these dynamic effects into account are explored. Two different levels of complexity are considered: we start with a laboratory setup and move towards a model of longitudinal thermoacoustic modes in a can combustor. First, state estimation for the electrically heated Rijke tube is considered. A globally convergent observer, taking into account nonlinearities from the electrical heater and distributed dynamics, is proposed and analysed. This is paired with a parameter identifier for estimating boundary acoustic impedances. The state observer and parameter identifier are tested both in simulations and experimentally. Next, a parameter identifier to estimate both boundary parameters of 2 × 2 linear hyperbolic systems with a single boundary measurement is proposed. Also, a transient model of acoustics in a duct with spatially varying cross-sectional area is derived. Using these two results together the boundary parameter estimation scheme for the Rijke tube is extended to more general ducts. An output feedback controller, combining a full-state feedback control law and collocated boundary observer, for longitudinal thermoacoustic instabilities in a model of a can combustor with distributed acoustics and a linear flame model is proposed next. Convergence is proven and it is tested in simulations. Lastly, the state estimation problem for a can combustor model with a nonlinear flame is considered. Neural networks are used to design an observer for the flame subsystem, which is subsequently verifed on CFD data.
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Monday, March 28, 2022 - 10:16:26 AM
Last modification on : Wednesday, March 30, 2022 - 3:06:49 AM
Long-term archiving on: : Wednesday, June 29, 2022 - 6:23:58 PM


Version validated by the jury (STAR)


  • HAL Id : tel-03621302, version 1


Nils Christian Aars Wilhelmsen. Estimation algorithms for thermoacoustic instabilities with distributed and nonlinear dynamics. Automatic Control Engineering. Université Paris sciences et lettres, 2021. English. ⟨NNT : 2021UPSLM055⟩. ⟨tel-03621302⟩



Record views


Files downloads